Architecture Overview
CapiBot is designed as a complete AI workforce platform. Understanding how the pieces connect helps you use it effectively.
The Big Picture
┌─────────────────────────────────────────────────────────────┐
│ YOU (Human User) │
└──────────────┬──────────────────────────────────────────────┘
│
┌──────────┼──────────┬──────────┐
│ │ │ │
▼ ▼ ▼ ▼
┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐
│Mission│ │Telegram│ │WhatsApp│ │ Slack │
│Control│ │ Bot │ │ Bot │ │ Bot │
│(Web UI)│ │ │ │ │ │ │
└───┬───┘ └───┬───┘ └───┬───┘ └───┬───┘
│ │ │ │
└──────────┴────┬─────┴──────────┘
│
┌────────▼────────┐
│ CapiBot Core │
│ (AI Engine) │
└────────┬────────┘
│
┌───────────┼───────────┐
│ │ │
▼ ▼ ▼
┌────────┐ ┌────────┐ ┌────────┐
│ Agents │ │ Tasks │ │ Memory │
│ (150+) │ │Kanban │ │ & │
│ │ │ Board │ │Knowledge│
└────────┘ └────────┘ └────────┘
Three Ways to Interact
1. Mission Control (Web Dashboard)
Your command center for managing everything:
- View all agents and their status
- Manage tasks on Kanban boards
- Monitor activity in real-time
- Configure settings and integrations
- Access memory and knowledge base
Best for: Day-to-day management, reviewing work, configuring the system
2. Messaging Apps (Telegram, WhatsApp, Slack)
Chat directly with your AI agents:
- Send messages and get responses
- Share files and documents
- Get voice message transcriptions
- Receive notifications
Best for: Quick interactions, on-the-go access, team collaboration
3. AI Companies (Autonomous Mode)
Let agents run independently:
- CEO agent creates business plans
- Teams self-organize and execute
- Human oversight when you want it
- Automated task delegation
Best for: Complex projects, ongoing operations, hands-off automation
Core Components
Agents
CapiBot comes with 150+ pre-built agents organized by specialty:
- CEOs — Strategic leaders who run companies
- Engineers — Build features and write code
- Designers — Create UI/UX and visuals
- Marketers — Handle campaigns and content
- Researchers — Analyze data and competitors
- And more...
Each agent has:
- A unique personality (SOUL.md)
- Specific capabilities
- Memory that persists across conversations
- Status tracking (idle, busy, offline)
Tasks
Work is organized as tasks that flow through a Kanban board:
- Inbox — New tasks waiting to be started
- In Progress — Currently being worked on
- Review — Ready for quality check
- Human Review — Needs your approval
- Blocked — Waiting on something
- Done — Completed
Memory & Knowledge
Your AI remembers everything:
- Agent Memory — Each agent's personal notes and context
- Knowledge Base — Shared information accessible to all agents
- Company Knowledge — Specific to each AI company
- Semantic Search — Find information by meaning, not just keywords
Channels
Connect through your preferred platform:
- Telegram — Full-featured bot with voice transcription
- WhatsApp — Business messaging with media support
- Slack — Thread-aware team collaboration
- Web Chat — Built into Mission Control
How Data Flows
Example: You Ask an Agent to Research Competitors
- You send a message via Telegram: "Research our top 5 competitors"
- Channel Service receives the message and routes it
- DeepAgent Service selects the Research Analyst agent
- Agent uses tools to search the web and analyze data
- Task is created and tracked on the Kanban board
- Results are stored in memory and sent back to you
- Activity is logged for your review
Example: AI Company Running a Marketing Campaign
- CEO Agent creates a business plan for the campaign
- You approve the plan in Mission Control
- CEO hires a Marketing Strategist and Content Writer
- Agents create tasks automatically
- Tasks move through the Kanban workflow
- Deliverables are saved to the company knowledge base
- You review and approve key decisions
When to Use What
Use Individual Agents When:
- You have a specific, one-time task
- You want direct control over the work
- You need quick answers or research
- You're exploring ideas
Example: "Atlas, write a blog post about AI trends"
Use Multiple Agents When:
- The task spans different specialties
- You want agents to collaborate
- You need diverse perspectives
- Work can happen in parallel
Example: Research Analyst gathers data, Content Writer creates copy, Designer makes visuals
Use AI Companies When:
- You have ongoing operations
- You want agents to self-organize
- You need consistent processes
- You want to scale without micromanaging
Example: A marketing company that runs monthly campaigns automatically
Use Subagents When:
- A task needs parallel execution
- You want to delegate sub-tasks
- The parent agent needs to focus on coordination
- Work can happen asynchronously
Example: CEO spawns 5 researchers to analyze different competitors simultaneously
Next Steps
- Learn about AI Companies and when to use them
- Understand the Agent Hierarchy
- See how Tasks flow through the system